KR20220143400A - Method for classifying cardiac arrhythmia from standard 12-lead Electrocardiogram signal using artificial neural network and classification device using the same - Google Patents

Abstract

The method of classifying cardiac arrhythmias from the standard 12-lead ECG signal is to pass the ECG signal of each of the 12 channels obtained through the standard 12-lead ECG through a bandpass filter in a predetermined range, then detect the R peak and calculate the detected R peak. R-R peak-based ECG sequence generation step to generate ECG sequences for each channel by grouping ECG signals to include three QRS complex waves as a reference A K-means clustering step of extracting a representative signal of each channel while dividing into a predetermined number of groups through K-means clustering, and a short time fourier (STFT) representative signal of each channel generated through the K-means clustering step transform) using a two-dimensional image extraction step and a two-dimensional image extraction step through STFT (Short time Fourier transform) for extracting a two-dimensional frequency map form including frequency change information according to time of a standard 12-lead ECG signal An artificial neural network to classify 8 types of cardiac arrhythmic diseases: atrial fibrillation, type 1 atrioventricular block, atrial premature contraction, ventricular premature contraction, left-brain block, right-angle block, ST-segment elevation, and ST-segment depression. It includes a cardiac arrhythmia classification step using an artificial neural network model to classify cardiac arrhythmias using the model input.

Description

Method for classifying cardiac arrhythmia from standard 12-lead Electrocardiogram signal using artificial neural network and classification device using the same

The present invention relates to a method for classifying cardiac arrhythmias from an electrocardiogram signal, and more particularly, to a method for classifying cardiac arrhythmias from a standard 12-lead electrocardiogram signal using an artificial neural network and a classification device using the same.

An algorithm for diagnosing cardiac arrhythmias from electrocardiogram signals has been proposed. These conventional algorithms classify cardiac arrhythmias by extracting time-frequency features from a single lead electrocardiogram signal. However, since some cardiac arrhythmias are observed only in a specific electrocardiogram channel, the classification accuracy depends on the type of cardiac arrhythmia to be classified. can be reduced.

Therefore, in order to accurately diagnose cardiac arrhythmias, it is necessary to comprehensively and closely check the ECG signals acquired from 12 channels. That is, the standard 12-lead ECG signal is used as an indicator for diagnosing cardiac arrhythmias because the shape of the signal changes due to changes in electrical and mechanical activity caused by various heart diseases. However, the symptoms observed in the electrocardiogram signal differ depending on the type of cardiac arrhythmia, and unlike persistent cardiac arrhythmias in which symptoms occur continuously and abnormal signs can be continuously observed in the electrocardiogram signal, single cardiac arrhythmias with intermittent symptoms In this case, it is difficult to visually determine the presence of cardiac arrhythmias from the ECG signals because abnormal signs are mixed between normal ECG signals. In addition, since the ECG signal measured through the standard 12-lead is different depending on the type of arrhythmia, it is necessary to analyze the ECG signals obtained from 12 channels in order to accurately determine the patient's cardiac arrhythmia.

In addition, prior studies do not differentiate between persistent cardiac arrhythmias and single cardiac arrhythmias, but focus on the characteristics of the electrocardiogram itself or classify arrhythmias through a feature engineering process for diagnosing specific arrhythmias. Since persistent cardiac arrhythmias and single cardiac arrhythmias have different characteristics in ECG signals, it is necessary to extract characteristics considering each characteristic for accurate diagnosis of cardiac arrhythmias.

S. Datta et al., "Identifying normal, AF and other abnormal ECG rhythms using a cascaded binary classifier," Comput. Cardiol. (2010)., vol. 44, pp. 1-4, 2017, doi: 10.22489/CinC.2017.173-154.

The present invention has been proposed to solve the above technical problems. Atrial fibrillation, type 1 atrioventricular block, atrial premature contraction, ventricular premature Provided are a method for classifying eight cardiac arrhythmia diseases, including contraction, left leg block, right leg block, ST-segment elevation, and ST-segment depression, and a classification device using the same.

According to an embodiment of the present invention for solving the above problem, the ECG signal of each of 12 channels obtained through a standard 12-lead electrocardiogram is passed through a bandpass filter in a predetermined range, and then the R peak is detected and the detected R The R-R peak-based ECG sequence generation step of generating the ECG sequence for each channel by grouping the ECG signals to include three QRS complex waves based on the peak, and the entire ECG sequence including the abnormal signs of single cardiac arrhythmias The K-means clustering step of extracting representative signals of each channel while dividing the electrocardiogram sequence into a predetermined number of groups through K-means clustering, and the K-means clustering step of extracting the representative signals of each channel generated through the A two-dimensional image extraction step through STFT (Short time Fourier transform) of extracting a two-dimensional frequency map form including frequency change information according to time of a standard 12-lead ECG signal using time fourier transform, and two-dimensional image extraction The two-dimensional frequency feature image extracted through the steps was used to classify eight cardiac arrhythmic diseases: atrial fibrillation, type 1 atrioventricular block, atrial premature contraction, ventricular premature contraction, left-brain block, right-angle block, ST-segment elevation, and ST-segment depression. A method for classifying cardiac arrhythmias from standard 12-lead electrocardiogram signals including the step of classifying cardiac arrhythmias using an artificial neural network model to classify cardiac arrhythmias using the artificial neural network model as input is provided.

In addition, in the R-R peak-based ECG sequence generation step included in the present invention, the R peak of each ECG signal is extracted by combining the Pan and Tomkins algorithm and the Robust thresholding algorithm.

In addition, in the artificial neural network model of the cardiac arrhythmia classification stage using the artificial neural network model included in the present invention, the two-dimensional feature image is extracted as an artificial intelligence feature while passing through the two-dimensional convolutional neural network layer. Finally, it is characterized by being classified as a cardiac arrhythmia disease.

As described above, through the present invention, it is possible to classify cardiac arrhythmias from standard 12-lead ECG signals, and it is possible to accurately detect some arrhythmia symptoms that are difficult to judge with the naked eye because they occur not continuously, but only once, through an artificial neural network model. It is possible to classify In addition, by analyzing the standard 12-lead ECG signals of patients who are expected to suffer from cardiac arrhythmias using the algorithm of the present invention, it may be used as an early diagnosis process before undergoing a detailed examination. The present invention can be used as an early diagnosis process before receiving a detailed diagnosis by doctors and experts after confirming the presence or absence of a heart disease in a medical center using a standard 12-guide electrocardiogram.

1 is a flowchart for a method for classifying cardiac arrhythmias from a standard 12-lead ECG signal and a flowchart for a clustering algorithm according to an embodiment of the present invention;
2 is an exemplary diagram of an RR peak-based sequence signal
3 is a diagram of a time-frequency characteristic map and an exemplary diagram of a time-frequency characteristic map generated in the proposed algorithm.
4 is a diagram showing a confusion matrix for the proposed model.
5 is a diagram showing a performance curve for the proposed model;

Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings in order to describe in detail enough that a person of ordinary skill in the art to which the present invention pertains can easily implement the technical idea of the present invention.

- 8 types of cardiac arrhythmias observed on standard 12-lead electrocardiogram

The electrocardiogram is widely used in diagnosing cardiac arrhythmias because the shape of the signal changes due to changes in the electrical and mechanical activity of the heart due to various heart diseases. The standard 12-lead ECG shows the electrical activity of the heart in four directions: Inferior (lead II, III, aVF), Anterior (lead I, aVL, V1, V2), Septal (V3, V4), and Lateral (V5, V6). According to the type of cardiac arrhythmia, the signal shape in each channel is different, and anomalies are sometimes observed only in a specific channel out of 12 channels. Among these cardiac arrhythmias, there are those in which abnormal signs are continuously observed on the electrocardiogram as a continuous symptom, and those in which abnormal signs are observed intermittently due to the occurrence of symptoms mixed with normal electrocardiogram signals. Therefore, it is not easy to judge cardiac arrhythmias by looking at the ECG signals with the naked eye. In order to accurately diagnose cardiac arrhythmias, it is necessary to closely and integratedly observe the ECG signals acquired from 12 channels.

- ECG signals due to persistent cardiac arrhythmias; AF, IAV-B, LBBB, RBBB

In the present invention, cardiac arrhythmias in which abnormal signs are continuously observed in the ECG signal are defined as 'persistent cardiac arrhythmias'. Representative persistent cardiac arrhythmias include atrial fibrillation (AF), fisrt-degree atrioventricular block (I-AVB), left bundle branch block (LBBB), and right bundle branch block (RBBB). In the ECG signal of AF patients, an intact QRS wave is observed, but due to multiple irregular electrical stimulation generated in the atrium, the P wave does not appear and irregularly vibrating fibrillatory waves are observed. Also, in the case of I-AVB, the PR interval of the electrocardiogram is continuously observed to be longer than that of the normal heart by 200 ms or more. In the case of LBBB and RBBB, anomalies are continuously observed, but the signs are not observed in all 12 channels. In the case of LBBB, an M-shaped signal is often observed in the V6 channel, and in the case of RBBB, an abnormality is observed in the V1 channel. However, it is not easy to distinguish these abnormal signs of persistent cardiac arrhythmias by motion noise, noise caused by electrode contact, and noise of the measuring device itself during the acquisition of the ECG signal.

- ECG signal due to single cardiac arrhythmias; STD, STE, PAC, PVC

Meanwhile, unlike persistent cardiac arrhythmias, cardiac arrhythmias occurring intermittently between normal signals were defined as 'episodic cardiac arrhythmias'. ST segment elevation (STE) and depression (STD), which are mainly caused by myocardial ischemia or infarction, are diagnosed when the height difference (ST segment slope) between the S peak of the ECG and the PR segment is ±0.5 mm or more. There is a characteristic that is observed unilaterally between the ECG waveforms. STD, in which the S peak is about 3 mm lower than that of the PR segment, is generally observed in leads I, V5, V6, and aVL, and normal ECG signals may also appear in V2-V3. On the other hand, in the case of STE, the S peak is observed 4 mm higher than in the PR segment. In the ischemic ST segment, the ECG signal in V2-V4 is almost normal, but in the case of non-ischemic ST segment, it is abnormal in V2-V3. Signs may also be observed. Premature atrial contraction (PAC) is a symptom that occurs when the tissue in the atrium is temporarily excited more than the threshold faster than the sinus node, and it is a common symptom that 80% of people suffer from a single episode during daily life. P wave is not observed in the ECG signal of patients with premature ventricular contraction (PVC), and the heart rate increases compared to a normal heart. 60-80 ms). However, it is not easy to accurately diagnose such a single cardiac arrhythmia because it is measured intermittently along with a normal electrocardiogram.

In order to accurately determine a patient's cardiac arrhythmias, the present invention proposes an algorithm and a cardiac arrhythmia classification device using the same for automatically extracting abnormal signs due to single cardiac arrhythmias, including persistent cardiac arrhythmias, from ECG signals obtained from a standard 12-lead electrocardiogram. . That is, the cardiac arrhythmia classifier receives an electrocardiogram signal from a standard 12-lead electrocardiogram, processes the signal using the proposed algorithm, and classifies the arrhythmia.

In addition, from this, eight cardiac arrhythmias - atrial fibrillation (AF), first-degree atrioventricular block (I-AVB), left bundle branch block (LBBB) and right bundle branch block; RBBB), premature atrial complex (PAC), premature ventricular complex (PVC), ST-segment depression (STD), and ST-segment elevation (STE)- An artificial neural network model for classification is proposed.

1 is a diagram illustrating a flowchart for a method for classifying cardiac arrhythmias from a standard 12-lead ECG signal and a flowchart for a clustering algorithm according to an embodiment of the present invention.

Referring to FIG. 1 , eight types of cardiac arrhythmias are classified through the following steps. First, the ECG signal obtained through a standard 12-lead ECG is sampled to have two intact ECG signals after extracting frequency signals between 8 and 16Hz using a bandpass filter. In some embodiments, a bandpass filter of 0.5 to 20 Hz may be applied to remove fluctuation noise and minimize data loss.

For this, R peak is detected based on the “Robust thresholding” algorithm, and then from the first detected R peak to the fourth detected R peak (PQRST signal of 3 cycles) is extracted. A K-means clustering algorithm is used to find representative signals that are judged to have signs of cardiac arrhythmias among the extracted signals. A representative signal is extracted by repeating the grouping of n signals extracted through the K-means clustering algorithm into two partitions.

The 12 signals finally detected from each of the 12 channels are converted into a two-dimensional image through short time Fourier transform (STFT). The converted 2D image includes information on frequency change over time of the standard 12-lead ECG signal, and information on specific cardiac arrhythmias is also included, so it is classified as one of 8 types of cardiac arrhythmias through an artificial neural network model.

Hereinafter, a method for classifying cardiac arrhythmias from a standard 12-lead electrocardiogram signal using an artificial neural network and a cardiac arrhythmia classification device using the same will be described in detail as follows.

First, the data used were standard 12-lead ECG signals of 6,877 (male: 3,699, female: 3,178) published in the Computing in Cardiology 2020 Physionet challenge. The ECG signal data used include ECG signals from patients with AF, including normal sinus rhythm without any cardiac arrhythmia disease, ECG from patients with I-AVB, ECG from patients with LBBB and RBBB, ECG from patients with PAC and PVC. Electrocardiography, STD and STE patients include electrocardiograms. Each signal was sampled every 500Hz, and each ECG signal was measured for a minimum of 6 seconds and a maximum of 60 seconds.

A bandpass filter was applied to remove the signal noise and baseline fluctuation noise from the original ECG signal, which includes motion sound and signal noise during measurement using an ECG device. In the present invention, a bandpass filter of 8 to 16 Hz or 0.5 to 20 Hz is applied to remove fluctuation noise and minimize data loss.

- R-R peak based sampling step (S10)

The R-R peak-based sampling step (S10), that is, the R-R peak-based ECG sequence generation step will be described in detail.

In the present invention, in order to classify eight cardiac arrhythmias by extracting the features of a sequence signal that includes both heart rate change and ECG signal shape information due to persistent cardiac arrhythmias, the R peak is detected and then the detected R peak is based on The ECG signals were grouped (creating an ECG sequence) so that three QRS complex waves were included. The R peak of each ECG signal was extracted by combining the Pan and Tomkins algorithm and the Robust thresholding algorithm.

That is, the Pan and Tompkins algorithm is used to extract the R peak from the signal of 8 ~ 16Hz extracted through the bandpass filter. After differentiating the filtered signal to extract the slope information of the signal, a normalization process is performed (d[n]). After squaring the normalized signal to increase the amplitude of the R peak signal (d ² [n]), the R peak is detected based on the “Robust thresholding” algorithm.

The “Robust thresholding” algorithm detects R peak by substituting 0 when the signal energy is less than the threshold parameter (η). At this time, the threshold value is adjusted according to each ECG segment.

At this time, the squared signal causes a problem of distorting the length of the QS interval of the electrocardiogram. To prevent this, the detection accuracy of the QRS wave is improved by using Shannon energy transformation. Shannon energy is calculated by normalizing the previously obtained threshold energy signal.

After the Shannon energy signal is refined through a moving average filter, a positive R peak and a negative R peak are detected through a first-order Gaussian filter.

Based on the R peak detected in the entire signal, the ECG signal is sampled from the first detected R peak (r1) to the fourth detected R peak (r4) so that the two ECG cycles are intact (electrocardiogram sequence signal generation) . Accordingly, the ECG sequence signal includes information on the heart rate calculated as the reciprocal of the R-R interval, including shape information of the PQRST waveform of the ECG.

2 is an exemplary diagram of an RR peak-based sequence signal, an example of an electrocardiogram signal sampled through such an algorithm. Depending on the type of ventricular arrhythmia and the characteristics of the patient, the sampled ECG signals do not all have the same number of samples. Therefore, to make each sample have the same number of sample data, the short sample data is augmented based on the longest sample data using linear interpolation.

- K-means clustering (cluster analysis) step (S20)

The signs of ECG signals due to cardiac arrhythmias may be observed continuously or may be observed temporarily between normal ECG signals. Therefore, in order to extract sample data containing abnormal signs due to cardiac arrhythmias, the ECG data is divided into several groups through K-means clustering. In this case, the operation of dividing the entire signal into two groups and dividing the signal into two groups again from fewer clusters is repeated until finally only one sample data signal remains. K-means clustering is performed for each of 12 channels, and representative signals are extracted from each of lead I, lead II, lead III, aVR, aVF, aVL, V1, V2, V3, V4, V5, and V6 channels.

For example, a clustering algorithm for extracting a representative signal including geometrical information due to a single cardiac arrhythmia from among the grouped signals is shown through the workflow of Fig. 1(b). First, a continuous signal group including three PQRST signals is grouped into two groups (A and B). In order to find a group of signals with abnormal symptoms due to single cardiac arrhythmias, if the ratio of the small group to the large group is 0.35 or less, one signal is randomly selected from the signals of the small group. One signal is randomly selected from the group's signals. If the number of divided two groups is the same, clustering is performed once more in each group and grouped into 4 groups (SA1, SA2, SB1, SB2), and one signal is randomly selected from the group with the smallest number among the 4 groups. pull out The extracted signal was used as a representative signal (S#), which included anomalies due to single cardiac arrhythmias and continuous cardiac arrhythmias.

- 2D image extraction step through STFT (S30)

Signals extracted from each channel are extracted as a two-dimensional array (in the form of a two-dimensional frequency map) through short time fourier transform (STFT) analysis to confirm a change in frequency over time. In this case, a linear interpolation or zero padding algorithm is used to match the sample data lengths extracted from each channel equally through K-means clustering.

3 is a diagram of a time-frequency characteristic map and an exemplary diagram of a time-frequency characteristic map generated in the proposed algorithm.

Referring to FIG. 3 , for signals extracted through STFT in each channel, only signals in the 0 to 50 Hz band, which are the main frequency band of the ECG signal, are extracted and then added below as shown in FIG. 3(a) to form one image. 3(a) and 3(b) are examples of time-dependent frequency feature images obtained from such an algorithm. The finally extracted frequency feature image includes information on changes in frequency over time in each channel and changes in signal frequency bands due to specific cardiac arrhythmias.

That is, representative signals extracted from each of the 12 channels are extracted as a frequency map according to time through a local Fourier transform (STFT). In the results, only frequency band information between 0 and 50 Hz was used (the frequency characteristic due to AF was observed in the range of 2.5-25 Hz), and time-frequency maps extracted from 12 channels were sequentially used (FIG. 3a). In the present invention, the frequency window was set to 300 Hz to observe the frequency change over time with a frequency resolution of 1.667 Hz, and the overlap ratio was set to 270 Hz to minimize the decrease in time resolution due to the increase in frequency resolution. Since the length of the R-R grouped signals is different depending on the type of cardiac arrhythmias, the time-frequency map extracted as a result of STFT was collectively converted into a 120×120 time-frequency map format using linear interpolation. Examples of time-frequency maps extracted through the algorithm used in the present invention are shown in FIGS. 3(b) and 3(c), which were used as input of a two-dimensional convolution artificial intelligence model.

- Cardiac arrhythmia classification step using artificial neural network model (S40)

The frequency feature images extracted through the above-described steps were obtained from 8 types of cardiac arrhythmic diseases: atrial fibrillation, type 1 atrioventricular block, atrial premature contraction, ventricular premature contraction, left-brain block, right-angle block, ST-segment elevation, and ST-segment depression. It is used as an input to the artificial neural network model for classification. In the artificial neural network model, two-dimensional feature images are extracted as artificial intelligence features while passing through a two-dimensional convolutional neural network layer, and then finally classified as cardiac arrhythmias. It is possible to classify cardiac arrhythmias more accurately through learning times, and the classification result can be confirmed relatively faster than to check cardiac arrhythmias by comparing 12-channel electrocardiograms visually.

That is, the CNN model for classifying cardiac arrhythmias consisted of three two-dimensional CNN layers, a BatchNormaliza-tion (BN) layer, and a two-dimensional Maxpooling (2DMaxP) layer (2DCNN-BN-2DMaxP). A dropout rate of 0.5 was applied to the first and second 2DCNN-BN-2DMaxP layers, respectively, and a dropout rate of 0.3 was applied to the last 2DCNN-BN-2DMaxP layer. Kernel regularization with loss_lambda of 0.00015 and bias regularization were applied to each 2D CNN layer to prevent the model from overfitting the training data. ReLU was used for the activation function of each layer, and softmax was used for the activation function of the output layer.

For model training, 80% of the total data was used as training data (5,484 pieces) and 20% as test data (1,372 pieces). The model was trained 500 times using the sparse categorical crossentropy loss function and the Adam optimization function with a learning rate of 0.01, and the optimal model was verified using 5-fold cross-validation. The classification performance of the final model was evaluated through accuracy, recall, precision, F1 score, confusion matrix, ROC curve, and precision-recall curve.

4 is a diagram illustrating a confusion matrix for the proposed model, and FIG. 5 is a diagram illustrating a performance curve for the proposed model.

4 and 5, the classification results of eight cardiac arrhythmias through the proposed algorithm are shown. The proposed model showed an F1 score of 0.8 or higher in the classification of AF and persistent cardiac arrhythmias such as I-AVB, LBBB, and RBBB. The F1 score was the highest in LBBB at 0.89, and the classification F1 scores in AF and RBBB were 0.86 and 0.85, respectively. The F1 score of I-AVB was 0.80. On the other hand, the classification performance of single cardiac arrhythmias such as PAC, PVC, and STE was relatively low. In particular, in the classification of STE, which had the smallest number of data classes, the F1 score was the lowest at 0.52 (the F1 score of PAC and PVC was 0.53, respectively). and 0.64). In addition, in the classification of Normal and STD, it showed moderate classification performance with F1 scores of 0.77 and 0.76. There was a difference in the number of data used according to each class. Accordingly, the final macro F1 score of the proposed model was 0.74, but the weighted F1 score considering the difference in the number of data for each class was 0.78.

To quantitatively evaluate the classification performance of the proposed model, two performance curves were used and compared (Fig. 5). Receiver operating characteristics (ROC) curves show the true positive rate according to the false positive rate of the classified data, and the classification accuracy of each class can be checked through the area under the curve (AUC). The AUCs of the persistent cardiac arrhythmias group such as AF, I-AVB, LBBB, and RBBB were 0.90, 0.90, 0.91, and 0.90, respectively, and the AUCs of Normal and STD were 0.87 and 0.89, respectively. However, the AUCs of the single cardiac arrhythmias group such as PAC, PVC, and STE were 0.72, 0.79, and 0.78, showing intermediate accuracy ( FIG. 5a - Receiver operating characteristics (ROC) curves -).

Cardiac arrhythmia data used in model training and classification differed in the number of data for each class. Accordingly, the classification performance of the model was evaluated through the precision-recall curve considering the difference in the number of data for each class. According to the classification accuracy, AF, LBBB, and RBBB showed excellent performance with AUC of 0.90 or higher among the persistent cardiac arrhythmias group (AUCs; 0.93 for AF, 0.94 for I-AVB, and 0.93 for RBBB), and I-AVB and Normal , showed a classification accuracy of 0.80 or higher in STD (AUCs; 0.88 for I-AVB, 0.85 for Normal, and 0.82 for STD). However, as in the ROC curve, the AUCs of PAC, PVC, and STE were 0.52, 0.68, and 0.55, respectively, showing low classification accuracy (Fig. 5b - Precision-recall curves - ).

As described above, the proposed method for classifying cardiac arrhythmias from standard 12-lead ECG signals is:

First, a bandpass filter of 8-16Hz is applied to remove meaningless noise and baseline wandering of the ECG signal from the ECG signal measured through a standard 12-lead ECG, and to extract only the significant ECG signal. The filtered time series signal is subjected to a process of dividing into a certain data length to learn an artificial intelligence algorithm using a short-duration ECG signal and to derive a cardiac arrhythmia classification result. At this time, the R peak is detected using the Pan and Tompkins algorithm and the Robust thres holding algorithm so that the divided ECG signal includes the PQRST waveform information and the heart rate information of the intact ECG, and from the first detected R peak A plurality of ECG sequence signals are generated by dividing the signal up to the fourth detected R peak. On the other hand, as mentioned above, the shape observed in the electrocardiogram signal varies depending on the type of cardiac arrhythmia. In particular, in the case of event-type cardiac arrhythmia disease in which symptoms occur unilaterally, an abnormal signal is observed only in a portion of the entire time series signal. . Therefore, in order to extract only the ECG signal sequences with anomalies of single cardiac arrhythmias, including anomalies of persistent cardiac arrhythmias, where symptoms can be continuously observed in the ECG signals, we use the K-means algorithm to extract the sequence signals with anomalies. and a normal sequence signal, and an ECG sequence determined to have an abnormality is selected as a representative signal of a specific ECG channel. The 12 signals finally selected from each of the 12 channels are transformed into a two-dimensional image form including frequency change information over time through short rune Fourier transform (STFI) analysis. The transformed two-dimensional image is classified as one of eight cardiac arrhythmias through a two-dimensional convolutional artificial intelligence model.

Through the present invention, it is possible to classify cardiac arrhythmia diseases from standard 12-lead ECG signals, and it is possible to accurately classify some arrhythmia symptoms that are difficult to judge with the naked eye because they occur not continuously rather than continuously, and are difficult to judge with the naked eye through an artificial neural network model. . In addition, by analyzing the standard 12-lead ECG signals of patients who are expected to suffer from cardiac arrhythmias using the algorithm of the present invention, it may be used as an early diagnosis process before undergoing a detailed examination. The present invention can be used as an early diagnosis process before receiving a detailed diagnosis by doctors and experts after confirming the presence or absence of a heart disease in a medical center using a standard 12-guide electrocardiogram.

As such, those skilled in the art to which the present invention pertains will understand that the present invention may be embodied in other specific forms without changing the technical spirit or essential characteristics thereof. Therefore, it should be understood that the embodiments described above are illustrative in all respects and not restrictive. The scope of the present invention is indicated by the following claims rather than the above detailed description, and all changes or modifications derived from the meaning and scope of the claims and their equivalent concepts should be interpreted as being included in the scope of the present invention. do.

Claims (4)

After passing the electrocardiogram signal of each of the 12 channels obtained through the standard 12-lead electrocardiograph through a bandpass filter in a predetermined range, the R peak is detected and the electrocardiogram signal is processed so that three QRS complex waves are included based on the detected R peak. RR peak-based ECG sequence generation step of grouping to generate an ECG sequence for each channel;
a K-means clustering step of extracting representative signals of each channel while dividing the entire ECG sequence into a predetermined number of groups through K-means clustering to extract an electrocardiogram sequence including abnormal signs of single cardiac arrhythmias;
STFT for extracting a two-dimensional frequency map form including time-dependent frequency change information of a standard 12-lead ECG signal using STFT (Short Time Fourier Transform) of a representative signal of each channel generated through the K-means clustering step (Short time fourier transform) a two-dimensional image extraction step; and
8 types of cardiac arrhythmias: atrial fibrillation, type 1 atrioventricular block, atrial premature contraction, ventricular premature contraction, left-angle block, right-angle block, ST-segment elevation, and ST-segment depression A cardiac arrhythmia classification step using an artificial neural network model for classifying cardiac arrhythmias using an artificial neural network model for classifying diseases;
A method of classifying cardiac arrhythmias from standard 12-lead ECG signals containing
According to claim 1,
In the step of generating the RR peak-based electrocardiogram sequence,
A method for classifying cardiac arrhythmias from standard 12-lead ECG signals, characterized in that the R peak of each ECG signal is extracted by combining the Pan and Tomkins algorithm and the Robust thresholding algorithm.
According to claim 1,
In the artificial neural network model of the cardiac arrhythmia classification stage using the artificial neural network model, the two-dimensional feature image is extracted as an artificial intelligence feature while passing through the 2D convolutional neural network layer, and finally, cardiac arrhythmia disease. A method of classifying cardiac arrhythmias from a standard 12-lead electrocardiogram signal, characterized in that they are classified.
A cardiac arrhythmia classifier using the method of any one of claims 1 to 3.

Images